BlackRock: AI agents could drive demand for stablecoins

BlackRock’s Digital Assets Research team published a whitepaper the week of Sept. 22, 2026, arguing that autonomous AI agents could generate new demand for stablecoins and tokenized claims on computing power. The paper, titled The Machine-Native Economy, was authored by Will Su, Robert Mitchnick, Jay Jacobs and William Helm and describes the market as nascent.

The note defines agentic AI systems as software that plans and executes multistep tasks with limited human oversight. It says such systems will need to pay for data, services and compute automatically, often in very small, high-frequency transactions. The authors contrast those needs with conventional payment rails — which require onboarding steps, slower settlement and fee structures — and point to stablecoins as a more suitable instrument because they provide a stable unit of account and around-the-clock settlement.

The paper cites market figures to support the payments case. It reports circulating stablecoin market capitalization above $300 billion as of September 2026 and adjusted stablecoin transaction volume above $11 trillion in 2025. The authors state stablecoin volume has compounded at roughly 80% annually since 2020, compared with about 8.5% for ACH; ACH moved about $93 trillion in 2025. The paper also notes an estimate that stablecoins represented roughly 3% of international payments.

Beyond payments, the research advances the idea of tokenized compute. The authors propose that standardized claims on computing capacity could be minted as tokens that are bought, sold, pledged as collateral and settled programmatically. In that model, autonomous agents would compare compute offers on price, performance, latency, location and hardware specialization, provision resources and settle payments without human intervention. The paper adds that compute-market liquidity is currently limited.

The whitepaper identifies several protocols and early efforts that could support machine commerce, including MCP and A2A for data and coordination and x402 for machine payments. An implementation of x402 that embeds payment for an API call inside the same HTTP request is cited as an early working example. The authors also reference ACP and Stripe’s machine-payments efforts and note that institutional settlement infrastructure is developing in parallel, including a chain recently opened by Circle with BlackRock among its validators.

The paper acknowledges that agent-driven payment activity is small today. Citing blockchain analytics data, the authors estimate AI agents account for between about 0.6% and 7.5% of payment volume on x402 and note most x402 traffic is not agentic. The report frames its claims as projections rather than established outcomes.

The authors point to two factors that could alter the outlook. If banks introduce programmable, tokenized deposit accounts, the exclusive role for stablecoins in machine settlements could weaken. Separately, regulatory proposals from the Federal Reserve on reserves, capital and custody for stablecoins could change how these instruments are issued and settled. The paper also notes that compliance frameworks designed for human-initiated payments may not map easily to continuous, sub-cent machine transfers.

The whitepaper summarizes the convergence with a single-line formulation: “AI represents machine-native intelligence, while digital assets represent machine-native money.” The authors repeatedly caution that the customer base they describe has not yet appeared in large numbers and that the ecosystem remains under development.